Artificial Analysis launches Text-to-Image Leaderboard & Arena on Hugging Face Spaces

AI Impact Summary

Artificial Analysis introduces a crowd-sourced leaderboard and image arena to rank Text-to-Image models using human preferences, producing ELO-scored comparisons from 45,000+ votes across more than 700 images per model. The effort surfaces relative strengths of leading proprietary and open-source models, including Midjourney, OpenAI DALL·E, Stable Diffusion 3, and Playground AI, with Stable Diffusion 3 Medium open-sourced on June 12. For technical teams, this benchmark provides a real-world, human-judged quality baseline to inform model selection, procurement, and integration timelines beyond automated metrics.